Research On Key Technologies Of Web Service Composition

The Number of network resources obtained on Internet is rapidly increasing, including storage resources, computing resources, application and software resources, which makes the Internet become an open distributed computing platform. Service-Oriented Computing (SOC) and Service-Oriented Architecture (SOA) are the latest development direction of distributed computing and software development. They can help enterprises or Internet users develop applications and business processes more flexibly, fast, and at lower cost. Among technologies that SOC used, Web service composition is the key technology.With the development of cloud computing, e-commerce and Internet technologies, the environment of Web service composition changes. In order to achieve universality, a number of Web service composition methods based on QoS are designed in view of the general environment, and do not take into account the specific environment and hence ignore some new problems and opportunities. This dissertation investigates these specific environments of Web services, and studies QoS-based Web service composition. We improve traditional ways to adapt to these specific environments, or put forward new methods to make full use of the advantages of these specific environments. Given the new problems brought by specific environments, we put forward the corresponding solutions. In this dissertation, main contents and contributions are as follows: 1. Internet is composed of subnets with different delays, thus for different users in different network locations, experiences on Web services are different. However, service providers only provide average QoS as a general evaluation index of a service, making most service composition methods do not take into account the current network transmission performance, which affect Web service invocations by different users. In addition, considering the Web service processing ability is limited, we use multiple candidate services to satisfy multiple service requests from different users. The advantage is that it not only improves the reliability of composite services and the ability of resisting disaster, but also significantly reduces the total processing time of composite Web services. However, research in this case is insufficient. Based on the queuing network model, this paper introduces the quantitative indicators of network QoS, points out the optimization variables, constraints and their limits. A runtime service composition method is proposed. First, this method use nonlinear optimization theory to choose a set of candidate services, then according to the actual status of the current network, select the appropriate candidate services at runtime. Experiments show that the proposed approach possesses the optimality and the execution efficiency, and it can adapt to the situation that parameters dynamicly change.2. The previous research focus on energy efficiency of atomic Web services or cluster services, but neglect the study of energy consumptions of composite services. Based on the energy consumption model of candidate Web services, this paper calculates the energy consumptions of the whole composite services, and it proposes an energy consumption awareness service composition method. This method contains three optimization goals, the first optimization goal is faster execution time:a combination method that can aware energy consumption should not greatly reduce the execution time to raise the utilization ratio of energy consumption. The second optimization goal is high reliability:some atomic services’service rate in a composite service is lower, but they have higher reliability, thus these services have advantages that making them together leads to higher overall reliability. The third optimization goal is low energy consumption:according to the energy consumption model to choose services, this paper uses M/M/c model that is closer to the actual situation, which reduces the overall energy consumptions of a composite service. To accelerate the solving speed, the optimal flow distribution is calculated by the hybrid algorithm in this paper. We conduct experiments in the different service scales (the number of abstract services and candidate services are different). The results show that this method, on the premise of meeting service requests, can greatly reduce the energy consumption of composite services.3. In the Internet environment, the Web service composition has two kinds of uncertainty, the first one is the uncertainty of the service invocation results, and the second one is the uncertainty of QoS. Using discrete time Markov Decision Process to model service composition, previous studies take the comprehensive value of each target as immediate payment, and then uses the optimal equation to solve the optimal strategy. However, the process of this kind of method needs to understand each state transition probability, and transition probability in the actual system is hard to obtain. In addition, previous research do not consider QoS value as a certain probability distribution. Based on the model of related literatures, this paper further expands the service composition model. Introducing QoS probability distribution to service composition model, this method uses machine learning algorithm, which is independent of models, to obtain optimal composite services. The experimental results show that this method’s learning cycle is short, and its adaptability is strong.4. Previous studies suggest that the relationships among service QoS properties are independent of each other, thus QoS attribute changes affect customer value inconspicuously. In this paper, our method considers the correlation of QoS attributes of a service, thus it can assign proper weights for the QoS to reflect service quality more accurately. In addition, matching between the enterprise business ability and IT level has been the highest priority of enterprise development, SOA is the key technology to promote the match. However, if no one can integrate strategic, tactical and operational level, the advantages of using SOA will have a hard time in the enterprise business layer. Considering the above two factors, this paper puts forward an improved Analytic Hierarchy Process (AHP) method for service composition. The method combines decisions in strategic, tactical and operational level, at the same time it considers the interdependent QoS attributes. It ranks composite services, and then chooses plans that are most suitable for enterprise strategy.